Create a function to iterate over columns and create a new column each iteration in R

user_19192959 0
Viewed 107

On occassion I get survey data with likert scale string items that I need to change to numeric in order to calculate basic descriptive statistics. In order to do this, I usually use the case_when function to create a new column for each item and assign each data point a numeric value. I am trying to write a function that can do this for many different columns all at once, so that I don't have to keep copy and pasting code. I am relatively new to this so any help would be appreciated:)

Here is what I have done previously in R:

#create data frame
df <- data.frame(v1 = c("Definitely True", "Somewhat True","Somewhat False","Definitely False"),
                 v2 = c("Definitely False","Somewhat False","Somewhat True","Definitely True"))

#Use case_when to add numeric columns to dataframe
df$v1n <- case_when((df$v1 == "Definitely True")==TRUE ~ "1",
                         (df$v1 == "Somewhat True")==TRUE ~ "2",
                         (df$v1 == "Somewhat False")==TRUE ~ "3",
                         (df$v1 == "Definitely False")==TRUE ~ "4")
df$v2n <- case_when((df$v2 == "Definitely True")==TRUE ~ "1",
                         (df$v2 == "Somewhat True")==TRUE ~ "2",
                         (df$v2 == "Somewhat False")==TRUE ~ "3",
                         (df$v2 == "Definitely False")==TRUE ~ "4")

This works if I want to replace each string value with a numeric value and overwrite data in the existing columns:

for(i in colnames(data_x)) {
  data_x[[i]] <- case_when((data_x[,i] == "Definitely True")==TRUE ~ "1",
                         (data_x[,i] == "Somewhat True")==TRUE ~ "2",
                         (data_x[,i] == "Somewhat False")==TRUE ~ "3",
                         (data_x[,i] == "Definitely False")==TRUE ~ "4")
}

But I would like to find a way to create a new column for each iteration as I did with the copy and paste version. Here is something I have tried but I haven't had any success. Any help on this would be appreciated.

for(i in colnames(df)) {
  df[[var[i]]] <- case_when((df[,i] == "Definitely True")==TRUE ~ "1",
                         (df[,i] == "Somewhat True")==TRUE ~ "2",
                         (df[,i] == "Somewhat False")==TRUE ~ "3",
                         (df[,i] == "Definitely False")==TRUE ~ "4")
}
case r for-loop
2 Answers

dplyr

df %>%
  mutate(across(v1:v2, ~ case_when(
    . == "Definitely True" ~ "1", 
    . == "Somewhat True" ~ "2", 
    . == "Somewhat False" ~ "3", 
    TRUE ~ "4"
    ), .names = "{.col}n")
  )
#                 v1               v2 v1n v2n
# 1  Definitely True Definitely False   1   4
# 2    Somewhat True   Somewhat False   2   3
# 3   Somewhat False    Somewhat True   3   2
# 4 Definitely False  Definitely True   4   1
  • across gives us the ability to do one thing across multiple columns. We can use v1:v2-syntax, or one of the other dplyr selector functions like matches, starts_with, etc.
  • the second argument for across here is a tilde-function (rlang-style), inside which . is replaced with the column data each iteration. For instance, the first time that this tilde-function is evaluated, the . references the vector df$v1.
  • because the default action of mutate(across(...)) will be to replace the columns, I add .names= to control the naming of the resulting data. This notation uses glue-syntax, where {.col} is replaced by the name of the column being evaluated in each iteration.

base R

I'll add the optional use of a lookup map.

lookup <- c("Definitely True" = "1", "Somewhat True" = "2", "Somewhat False" = "3", "Definitely False" = "4")
df <- cbind(df, setNames(lapply(df[,1:2], function(z) lookup[z]), paste0(names(df[,1:2]), "n")))
rownames(df) <- NULL
df
#                 v1               v2 v1n v2n
# 1  Definitely True Definitely False   1   4
# 2    Somewhat True   Somewhat False   2   3
# 3   Somewhat False    Somewhat True   3   2
# 4 Definitely False  Definitely True   4   1
user_33582720

I'd be inclined to do this differently. If you convert the Likert Scale columns to factor, with levels in the correct order, you can use as.integer(...) to get the numeric levels directly, without all this case_when(...) business.

Here's an example using data.table

library(data.table)
likertScale <- c("Definitely True", "Somewhat True","Somewhat False","Definitely False")
cols        <- names(df)
setDT(df)[, c(cols):=lapply(.SD, factor, levels=likertScale)]
df[, paste0(cols, 'n'):=lapply(.SD, as.integer), .SDcols=cols]
df
##                  v1               v2 v1n v2n
## 1:  Definitely True Definitely False   1   4
## 2:    Somewhat True   Somewhat False   2   3
## 3:   Somewhat False    Somewhat True   3   2
## 4: Definitely False  Definitely True   4   1
user_29850070
Related